haha. true. :D

That's what HFT is doing as well, just on a faster & smaller scale. One of the results is that markets settle to the new clearing price much more quickly.

No, that's not HFT, that's front running, which is illegal (at least if you're a broker). Front runners may _use_ HFT, but it's not HFT. Chances are when you buy/sell your 15.3 shares of IBM through Schwab, the broker is either handling the trade using its own internal pool of commonly held stocks, or it's bundling your trade together with a bunch of others and using the HFT network to execute.

HFT is, at bottom, merely a logical extension of the numerous methods of increasing the speed and accuracy of trades, starting before the invention of the ticker tape in the late 1800s. Various entities use it in different ways. Back in the day, traders who paid for a ticker machine in their office were able to trade on news perhaps an hour earlier than those who didn't. Nothing has changed, only the scale.

I'll just add that I know of people who started out with their own little neural network program on their home PC, watching 15 minute delayed quotes from Yahoo (or in some cases just watching daily closings), and let their little PC grind away and first, recommend, then later execute trades, and made money - in some cases lots of money.

The point is it's a scale free network. You don't need to know what the HFT traders need to know - you only need to know what's relevant at your time and volume scale. The HFTs are trading on ripples that, if it were an ocean, would be far beyond the visible, and in so doing are making the market an order of magnitude more liquid and more 'efficient' (in the economics sense, as well as the general sense). You can profitably trade on ripples that are (by way of analogy) the size of your fingernail up to the length of your arm (days to months); and other people/entities can make money trading on the huge macroeconomic trends over years, like Berkshire Hathaway. (When the dotcom bubble was building, BH stocks were down to about $14,000 per share but I was too broke to buy any. Now it's something like $175,000 per share.)

Sure it is. The HFT market is now essentially how the great majority of retail trades are also accomplished - your broker either sesttles your trade internally, or bundles it in with thousands of others to make a single large trade. And your broker does have access to that information, which optimizes his price to you. You are free to get that information as well, of course - but it will cost much more than any trading you're likely to do. It's pretty much wholesale vs. retail.

No single entity (with the possible future exception of NSA...) has access to _all_ information. But if the players at a given level of activity have equal opportunities to acquire that information, then that works fine. I don't want to make this statement mean too much, so let's use Warren Buffett and Charley Munger as an example. These two folks pay zero attention to the day-to-day flips and flops of the market, because that's not their game. They play a 10-15 year game. Day traders with any sense are these days not playing a 10-second game, because that's the game of the big players who are paying $1000s per day for market information that is relevant on a sub-second level. Day traders should really be thinking in terms of at least days, preferably weeks or months, because the HFT folks have almost zero interest in anything beyond tomorrow.

IOW it's a scale-free system. There are ripples of all sizes from 'quantum fluctuations' - random fluctuations around the minimum price difference - +/-1 cent or 0.1 cent are the most probable, trailing out in an inverse power law relationship to some large number, which amounts to an equivalent to a tidal wave - but in perhaps 10 to the 10th dimensions (every possible connection between every possible market input). And the key is equal access to the _relevant_ information for all players _at that level_. Given that, the market will always tend to settle to an equilibrium state (though it never will reach it).

Interesting idea.

Hmm. I'd argue that computers are inherently rational, _within the limits of their programming_. Adaptive systems like neural networks are arguably more so in one sense, and less in another - depends on whether one considers rationality == deterministic. But studies have shown that people, especially men, are very emotional in their market decision making - our egos get in the way, and we fall into various bias traps.

Heck, I'm looking at a startup now that may well be a loser, but I 'like' the idea behind it. Once I'm convinced it takes a lot of negative information to change my mind - just like everyone else.

From my somewhat closer perspective (I'm involved in some financial stuff, and one of my cohorts is a risk manager at a major investment bank), I'd say that by and large the people/entities very much DO pay the price when things go wrong, nearly all the time. This applies to individual traders and to the institutions where they work. Even the ones you hear about are more often about the price being paid by the error-maker than about getting off scot free.

I am not a fan of this 'too big to fail' B.S. and bailouts in general - that is an example of how the big institutions have managed to successfully manipulate the political system and built various 'rent-seeking' mechanisms into the rules. But (OTOH) I do see where bailouts to some extent could be used to help protect the minor players - individual investors and pension funds, for example - from the errors. That is an area of very strong debate in all arenas from academia to the companies, to the legislatures. (Chances are that any legislation will be worse than the disease, but that's another topic.) But it's well documented that bailouts and insurance have a perverse effect of encouraging investors (and whoever) to do less due diligence, and invest in things that are a bad idea. I'm trying to think of the term for this but drawing a blank.

It's exactly the same thing as when one of my neighbors has a TV they want to sell for $10, and I buy it and sell it to a friend across the river for $11. In both cases everyone is satisfied with the value given for value received. Without my knowledge the neighbor might not have sold it at all. I may or may not know what my friend would actually pay, so I may be betting that he will pay $11 (if I do know for sure, it's more arbitrage, if not, it's more speculation). And I also paid for the phone call out of my $1 margin. There are complications we can discuss such as borrowing the money to buy the TV long enough to get paid by my friend, but that's what it is. Arbitrage is nothing more than balancing the difference between two markets.

It's also exactly the same thing as a river between two lakes - the river allows the two lakes to reach a common level, at a rate determined by the capacity of the river and various fluid dynamics considerations. Regulation has an important purpose here, such as setting the maximum flow rate to avoid erosion, flooding, 'tidal' waves splashing up on the opposite shore, etc.

There are ways within this system to cheat, but arbitrage per se is not cheating, it's implementing the system of equilibrating the markets.

Aha - this is the key insight! :) - at least as far as my intellectual model of markets (and economies) perceives. It is quite reasonable to use the physics of wave models on markets, increasingly so as computers take over the plumbing. For simple economic models, you can use the calculus of electronics and not go far wrong. A sudden event, like a sell of 100MM shares of X, looks to the market like a step function, and it will respond with a very rapid (near vertical) drop, which will 'ring' as it overshoots the new equilibrium, then reacts - a classic square wave without enough low pass filtering (if I get my low/high stuff right). HFT greatly improves that response with a much better & faster settling time to the new equilibrium.

IMHO, any market is an approximation of a 'complex adaptive system' (essentially any living system - ecosystems, brains, economies, political systems, and high school social hierarchies are some examples). These in turn are approximations that lie somewhere between 'billiard ball physics' and fluid dynamics. (For an illuminating experience, read "The Edge of Chaos" - a controversial but nevertheless very instructive book.) The finer the granulation of possible trades, the more closely the markets approximate a wave equation in a large number of dimensions, and the less the effect of 'quantization error' - fluctuations around the minimum price & trade execution time.

Ane can easily see the pool of common stocks that every broker maintains on their own books as analogous (or equivalent to) capacitors, or lakes, if one prefers to work with water. HFT has reduced the resistance (and impedance? I think so but IANA EE) in the wires, or increased the flow rate through the rivers and ocean currents if you prefer - it's reduced the viscosity of the fluid (money). The biggest difference to electronics and similar domains is the number of dimensions = it's like a circuit with 100,000,000,000 active elements and a similar number of interconnects - or more likely, another 5 orders of magnitude number of interconnects. That's why neural networks are very good at this stuff.

Sorry, I know I'm jumping around a lot - this was going to be my thesis topic in an economics PhD but I went another direction and never started the PhD program.

(The number of dimensions is an interesting area of exploration - it can be viewed in a variety of ways.)

WRT high school social hierarchies - research has shown that popularity and who-knows-who form an inverse power law distribution, which is indicative of a scale-free system or C.A.S. - a student's popularity is 1/N where N is the number of people who 'know' them.

There were citations to that earlier in the topic - I'm at work now so can't go back and look them up.

And agreed, the wave (it's still there) would be much slower and smaller. One of the relatively unknown bits about much legislation is what they call 'transition rules' that establish the rate at which changes are to take place. This often includes a gradual increase in a tax to the final amount. As a rather messy case in point, the Obamacare legislation takes a good part of a decade to go completely into effect.

As I noted, perhaps in another comment, I don't have a strong opinion on the actual tax proposed - I've been responding mainly to various commenters' lack of knowledge (excepting yourself, of course! :D ) Trading fees are so much smaller for everyone now that a small tax would not seem to make much of a difference - but it should be imposed gradually if large, and should not change back and forth, effectively 'whipping' the market response. I don't think that the government could successfully use the method of 'seeing how the market responds', unless the congress sets a maximum value and lets the SEC handle the implementation as a regulatory activity, and in fact I dislike it as a market control entirely (such a control can not respond quickly enough), though I'm perfectly OK with it as a money raising method to pay for regulation - as such they should just pick a rate and stick with it. Generally, when the government intervenes in a market - such as the Fed action, and the 'stimulus' - by the time the government's response goes into effect the market has already gone past the point where the effect would be as desired, and instead the government 'pumps the swing' up the other side, exacerbating the market swings.

Nothing to do with HFT, any more than bank robbing is a function of automobiles. Automobiles just made getting away faster. And, while it didn't get covered in the news, HFT also compensated for this glitch within minutes, rapidly moving markets back toward equilibrium.

Yep. :)

If JQP is attempting to do HFT, he's a sucker. It's a scale-free system, there are opportunities to profit (or lose...) at all frequencies. JQP needs to work at the time/price range that he is competent. 10 minutes is not the proper time/price range.

Manipulations such as trade cancellation are a separate problem. It's always been there, just faster now. (Heck, IIRC that's one of the ways Joseph Kennedy made his wad - and why Roosevelet made him head of the SEC, since he knew how to cheat he knew where to look for cheaters.) And regulators continue to work on those problems. The regulators themselves are now running HFT-like systems to monitor HFT-domain issues. Which is as it should be.

Your analogy neglects the difference between the roads used by a Model T and the modern roads. And blindfolded is inapplicable. A better analogy would be between a 747 and a Model T, but that has problems as well. The best analogy would be the most direct - between the Baron von Rothschild's messengers returning to London from Waterloo a few hours before the royal messengers, allowing him to arbitrage the stock market, and the ticker tape, allowing everyone to get the same information at the same time, much more quickly. The ticker tape revolutionized stock trading in the late 1800s, dramatically decreasing the gap between stock prices in different locales. HFT is just an extrapolation of that. Now the bubble is passing, it's just the way most trades are done, even when a retail trader sells 10 shares of IBM - that 10 shares is either handled internally by his broker, or bundled into a larger trade that goes through the HFT network.

(not to say I'm against the tax per se - I have no opinion on that.)

Having said that, I haven't said anything about whether the tax is a good idea or not. I really don't have much of an opinion, as long as it is small enough, and left in place unchanging for a long time. The price of trades is so much smaller for everyone than it used to be, so increasing that cost a small amount may not have much of an effect. It might be easier to just make the tax a percentage of the trading fee. Some form of tax may be a good way to raise revenue to pay for regulation and enforcement. But I'm not really up on what the taxes are now in different jurisdictions.

All of the research shows the opposite to be the case.

This is another topic - the question of using taxation as an implement of social policy. Experts over 100 years, from SCOTUS (Chief Justice Robert Taft was the first) to academic research have shown this to be a very bad idea for a number of reasons, including inefficiency, unintended consequences and affecting all the wrong people/entities.

Using your method (slowly increasing, ...) is making the government a participant in the market rather than a regulator, which is even worse. You're basically trying to establish the 'clearing price' of the tax, or if you prefer, trying to establish the most profitable supply/demand price - it's just like Levi's adjusting the price of their jeans to maximize profits. Proper regulation (even using a fixed tax, if it stays in place long enough) works by constructing the boundaries of the flows in the market(s) - think of them as the continents, islands, channels, jetties, etc. that establish the flows of the liquidity. But when governments participate in markets, they are no longer establishing boundaries, they are acting more like fish and boats, or whatever. Now every movement of the government, or even every signal that the government might do something, causes waves in the market - the opposite of what is desired. And the market response will now be casting that boat or fish about, affecting everything from the value of the dollar (or whatever currency) to unemployment in ways that are hard/impossible to predict.

Because that's the opposite of equilibration.

Without beating the horse more than I already have in a dozen other posts (I _really_ need to get to work!), HFT is basically implementing the interface between what one might call the 'quantum level' and the macro level. Now that the tech bubble aspect of HFT is popped and dissipating, and for every HFT algorithm there is another HFT algorithm second guessing that algorithm, HFT's primary effect is to damp out the small ripples in the global markets. Yes there will be random fluctuations on the order of T/N where T is the number of time quanta and N is the number of minimum price differences in a market (the probability of a price difference of N cents occuring increases over T microseconds). So the markets are now more purely scale-free and operate at a much finer level of granularity than used to be - it used to be minutes and eighths of a dollar, now it's microseconds and fractions of a cent. That's all. For everyone except HFT algorithms, the market is now just more efficient and the information more perfect.

Interestingly, the chaos is also self-damping as other HFT algorithms pick up on the distortion and work the other side of it. After a couple of 'rings' the signal is gone.